Addressing the Livelihood Crisis of FarmersWeather & Climate Services

WMO

Short, Medium and Extended Range Forecasting

B. Mukhopadhyay

Overview…………..

• Relevance of Weather prediction to agriculture

• Spatial and Temporal Scales of weather phenomena

• Predictability

• Conventional methods of forecast

• Global and Regional NWP Models

• Climate models

• Integrated Agro advisory services

Relevance to the Livelihood Issue

Yield without technology

AgrometTechnology

Ideal Yield

Yield aided by technology

Yield aided by agro-meteorologyNatural

Realization of the promise

A View From Space -hierarchical structure of planetary weather

Spatial Scales of Motion

Distribution of atmospheric moisture

JUNE MAY

2 May

seasonal transition

16 May

31 May

14 Jun

30 Jun

Cumulonimbus clouds can coalesce to form a huge continuous cloud cell, as seen in this meteosat image of the Balkans

View of an incoming storm low approaching the coast of California taken by SeaWIFS instrument on the OrbView-2 satellite.

Smaller spatial scales

Mix of fog and smog in northern India seen by MODIS

View of an "anvil cloud". from the International Space Station

Horizontal scale

LifeTime

Dust DevilTornado

Thunderstorm

Off-shore vortex

Meso-cyclone

Monsoon depression

TropicalCyclone

Extra-tropical Cyclone

Monsoon circulation

10 m/10min

10,000 km/100 days

Weather Systems:Temporal-Spatial Scales

Phenomena Geographical occurrence

Temporal /Spatial scale

Madden Julian Oscillation

Tropical 30 – 40 daysWave Number 2-4

Planetary waves Mid latitudes 10 – 20 daysBaroclinic waves Midlatitudes, subtropics - 3 daysTropical Cyclones Tropical - 7 daysEasterly waves Tropical - 7 daysThunderstorms - 5 hrsFog hrs - daysDroughts Any where Days – Months - yearsHeat / cold waves Except Equatorial regions Days - weeks

Very Limited Skill in Extended range time scale

(10 days to one month)

Time Series of Solution

-30

-20

-10

0

10

20

30

Time t

Variable X

Predictability & Forecasting RangePredictability & Forecasting Range

unpredictablepredictable

Predictability of the First Kind

Lorenz systemwith slight different initials

Tim e Series of Solution X

-30

-20

-10

0

10

20

30

0 500 1000 1500 2000

Tim e t

variable X

Time Series of Solution

-30

-20

-10

0

10

20

30

Time t

Variable X

Predictability of the Second Kind

LRF targets this signal

Lorenz systemwith a forcing

Major forcing is boundary conditions

Forecasting RangesRange Time Methods Phenomena Utility

Nowcasting - 6hrs Radars etc Hailstorms, Squalls with high accuracy

Severe Weather Warnings ( ~ 500 m)

Short Range 2-3 days Nested AtmosphericModels

Synoptic scale weather systems

Conventional Forecasting resolution (3-25 km)

Medium Range

7-10 days Global Atmospheric Models

Synoptic scale weather systems

Conventional Forecasting resolution(25 – 50 km)

Extended Range

10–30 days

Coupled Atmospheric

Persistent systems• Blocking Highs• MJO• ITCZ

Droughts and Heat / Cold Waves

Resolution vs Lead time

Forecasting by Synoptic Charts

PREPBUFR non-conventional

SOURCE: GTS SOURCE: ftp

GSI Analysis

FCST

06 hourly FCST

FCST7 days

Fixed Fields(climatology)

GLOBAL DATA ASSIMILATION (GSI)

00Z, 12Z

Obs/BG Error StatsUsage Control Coeff tablesLand surface

ShortShort--range Forecasting Strategy range Forecasting Strategy

• Triple nested model forecast27, 9 and 3 km

• Cycling4 times a day

Analysis

00Z 06ZTime

Model StateForecast

Global Model WRF-Var

ObservationsBE ModelObs. Error

Model

WRF Model

Model StateForecast

Global Model WRF-Var

ObservationsBE ModelObs. Error

Model

WRF Model

Analysis

ARPS & ADAS –

• Advance Regional Prediction System Model (ARPS) - has been used.•ADAS has been used to assimilate & interpolate, radar data file generated by 88D2ARPS & background and boundary file generated by EXT2ARPS, onto the ARPS grid using GFS Forecast.

• Following Diagram is showing ½ hourly assimilation cycle ( first 3 hours) & then 21 hours ARPS Model forecast -

9-km assimilation:GFS model provide background and boundary conditions

ADAS

0000Z 27 NOV 08 0100Z 0200Z 0300Z 0130Z

forecast

0030Z 0230Z

ADASforecast

ADASforecast

ADASforecast

ADASforecast

ADASforecast

ADASForecast (21 hrs)

0000Z28 NOV 08

IMDS.20081127.0000 IMDS.20081127.0300

For Cyclone Case

Generation of Multi-model Forecasts

Forecast (F)= WiFi

NCEPNCEP JMAJMA ECMWFECMWF NCMRWFNCMRWF UKMOUKMO

kjiW ,, ∑=

= 5

1,,

,,

kkji

kji

C

C

, i = 1, 2, ….., 161; j = 1,2,....,161

Parameters:RainfallMax and Min temperature

Total cloud cover Surface Relative humidity Surface Wind

IMD Multi-model Ensemble (MME) based District level (50 x 50 km) Forecasts

Model Performance

Spatial Distribution RainfallRoot Mean Square ErrorsMean Errors

Categorical Skill scores PODTSBS

Inter-comparison : Day-5 forecast of Cum. Rainfall (cm)

OBS

MME ENSM

Inter-comparison of day-5 forecast of Cum. Rainfall (cm)

ECMF UKMO

JMANCMR

NCEP

OBS

RMSE 2009 :Day-1

ECMF NCMRNCEPUKMO

JMA MME ENSM

RMSE 2009 :Day-5

ECMF NCMRUKMO NCEP

JMA MME ENSM

RMSE ERROR

11

12

13

14

15

16

17

DAY-1 DAY-2 DAY-3 DAY-4 DAY-5

RA

INFA

LL (m

m/d

ay)

GFS

ECMF

JMA

T254

UKM

MME

ENSM

Inter-comparison of model performance: RMSE 2009

Mean Errors (F-O) :Day-5

ECMF NCMRNCEPUKMO

JMA MME ENSM

TS : All India : Day-5

0

0.1

0.2

0.3

0.4

0.5

0.6

0.1 2.5 5 10 15 20 25 30 35 65 mm

Rainfall Threshold in mm

TS

GFSECMFJMAT254UKMOMMEENSM

Hit Rate : All India : Day-5

0

0.2

0.4

0.6

0.8

1

0.1 2.5 5 10 15 20 25 30 35 65 mm

Rainfall Threshold in mm

Hit

Rat

e

GFS

ECMFJMAT254

UKMOMMEENSM

Bias : All India : Day-5

0

0.5

1

1.5

2

2.5

0.1 2.5 5 10 15 20 25 30 35 65 mm

Rainfall Threshold in mm

Bia

s

GFSECMF

JMAT254

UKMOMME

ENSM

Bias : All India : Day-4

0

0.5

1

1.5

2

2.5

0.1 2.5 5 10 15 20 25 30 35 65 mm

Rainfall Threshold in mm

Bia

s

GFS

ECMFJMAT254UKMOMME

ENSM

Intra-seasonal Variability of Monsoon type systems-prediction of 15 days to 1 month

The active/break cycles -manifestations of sub-seasonal fluctuations of the northward propagating ITCZ.

Since the time scale is sufficiently long so that much of the memory of the atmospheric initial conditions is lost, and it is probably too short so that the variability of the ocean is not large enough, which makes it difficult to beat persistence.

However, an important source of predictability at this time range is the Madden–Julian oscillation (MJO).

• Intraseasonal Time Scale: ~40-60 days• Planetary-Scale: Zonal Wavenumbers 1-3• Baroclinic Wind Structure• Eastward Propagation

E. Hem: ~5 m/s, Surf.+Conv.+Circ. InteractionsW. Hem: ~ > 10 m/s, ~Free Tropospheric Wave

• Tendency to be Equatorially Trapped• Strong Seasonal Dependence:

NH Winter: Eastward Propagation NH Summer: ~Northeast Propagation

• Significant Interannual Variability• Potential Role of Ocean/SST Feedback• Convection Has Multi-Scale Structure• Significant Remote and Extra-Tropical Impacts

U200

U850

CloudyLow OLR

ClearHigh OLR

Rainfall

Typical Variables Used for MJO Analysis

MaddenMadden--Julian Oscillation (MJO)Julian Oscillation (MJO)

Experimental Forecast for two weeks (Empirical models)

1) Real-time Multivariate MJO Index – for 15 days Based on the first two Empirical Orthogonal Functions (EOFs) of the combined fields of near-equatorially-averaged 850 hPa zonal wind, 200 hPa zonal wind, and satellite-observed outgoing longwave radiation (OLR) data. Wheeler and Hendon (2004)).

2) US CLIVAR MJO Forecast Project - Available through Climate Prediction Centre, CPC).

3) MJO forecast in the form of OLR anomalies (a) Xavier and Goswami (1997) – Uses the analog method for the forecasting

of OLR pentad anomalies at lead time period of 1 pentad to 4 pentads(b) Jones et al., (2004) - The model uses principal components (PC’s) of

empirical orthogonal function analysis of 20-90 days anomalies of OLR. (Jones et al. 2004 J. Climate).

4) Self Organising Map - Developed by IITM is based on a non-linear pattern recognition technique known as Self Organising Map (SOM), which predicts rainfall using dynamical indices. Currently upgraded and used real time rainfall as input in the forecasting of probability of rainfall anomaly over central India. (IMD and IITM presently working together)

1) ECMWF Monthly forecast 51 ensembles - (days 5 to days 32) –Forecast in four weekly averages. (Runs once a week)

2) JMA/Tokyo Climate Center 51 ensembles – (day 2 to day 29) -Forecast in two weekly averages and one fortnightly averaged. (Runs once a week).

3) NCEP CFS 4 ensembles - Forecast as daily averages (Runs daily). Model runs every day with 4 ensemble members with 2 days lag of Ocean initial

condition

4) Experimental Multi-model forecast system for India : For two weeks after bias correction, based on ECMWF and NCEP coupled model

Experimental Forecast for two weeks-1 month(Dynamical Coupled models)

ECMWF forecast for Dry spell of September, 2009Valid for 14-20 Sep Valid for 21-27 Sep

Lag Week 1

Lag Week 2

Lag Week 3

Experimental Monthly Probability ForecastBased on Real time NCEP coupled model outputs

Total 60 forecast ensemble members

4 Ensembles/day

with 15 days of the Month (0)

Model Hindcast Climatology from 25 years (1981-2005)

Probability forecast (in %) for month 1

(i) Above Normal

(ii) Below Normal

Forecast (Tmax & Tmin) For March 2010 Based on February ICs

Observed Tmax anomaly, Mar 2010

Doppler Weather Radars

DWR at VSK

• Wind information• Water content in

clouds in different phases

• Digital output

Increases forecast accuracy dramatically over the next few hours and appreciably over 24 HRS

… Means of Nowcasting and very short range forecasting

Nowcasting

24 HR forecast using DWR data

1 HR 2 HR

1 HR 2 HR

Without DWR

With DWR

( Regions likely to experience Stem borer incidence on Rice crop)

Districts likely to be affected by the incidence of tikkadisease on Groundnut

Irrigation Requirement at different districts

Ratnagiri

Kolhapur

Dhule

Raigarh

Satara

Pune

Sangli

Solapur

Osmanabad

Mumbai

Thane

Ahmednagar

Sindhudurg

Nashik

Bid

JalnaAurangabad

Jalgaon

Buldana

Latur

Parbhani

Nanded

Akola

Yavatmal

Amravati

Gadchiroli

W ardha

Nagpur

Chandrapur

Bhandara

0 40 80 120 160 200

Kilom e te rs

.

Light Rainfall

Excess Rainfall causing water logging/Runoff

Punjab

No Rain

Postpone fertilizer

application at Gurdaspur

and Patiala due to excess

rainfall .

Farmers are advised to

apply fertilizers in the

remaining districts of the

state due to light rainfall

or no rain in these areas

Rainfall

> 20 knots (Squall)

> 34 Knots (Gale)

< 10 knots

Wind SpeedPunjab

In  Bhatinda and  Patiala    wind 

speed    is greater than 20 knots and 

in Jalandhar and Ropar wind speed   

is  greater  than  34  knots,  avoid 

applying  fertilizer  to  the  crop  in 

these districts.

Apply  the  necessary  fertilizer  in 

the  remaining  districts  of  the  state 

as  wind is calm .

Temp. > 28OC

Temp. < 28OC

PUNJAB

Apply light irrigation in Jalander and Ludhiana where temperature reaches more than 28OC.

Grape cultivated area in Maharashtra

Grape is grown in areas where the maximum temperature range is from 15-40°C. High temperatures above 40°C during the fruit growth and development

reduce fruit set and consequently the berry size.Minimum temperature >10°C is good for grapes

When normal maximum temperature of a station is less than or equal to 400C

When normal maximum temperature of a station is more than 400C

Apply frequent irrigations to the crops.Apply mulches to maintain high moisture status in the soil

In rabi season the most appropriate time of sowing for wheat crop is when the daily ambient temperature drops to 20-220C.

In Amritsar, Firozpur, Faridkot, Bathinda and Muktsar the temperature range is between 20 to 220C in the third week of November . In these districts, farmers can start sowing wheat. In the other districts, they can wait for sowing.

Bathinda220C

Muktsar200C

Firozpur21.2 0C

Agromet Product for Sowing of Crops

In rabi season the most appropriate time of sowing for wheat crop is when the daily ambient temperature drops to 20-220C.

In Amritsar, Firozpur, Faridkot, Bathinda and Muktsar the temperature range is between 20 to 220C in the third week of November . In these districts, farmers can start sowing wheat. In the other districts, they can wait for sowing.

Amritsar210C

Bathinda220C

Faridkot

210CMuktsar

200C

Firozpur21.2 0C

Agromet Product for Sowing of Crops

P/PET ≥ 0.5

P/PET < 0.5

0.35

0.3

0.45

0.5

0..5

0.5

0.5

P = Mean weekly rainfall (mm)

PE = Weekly potential evapotranspiration(mm)

Punjab For Kharif season• Monsoon arrived in Punjab in the first week of July.

• In Firozpur, Amritsar, Faridkot, Muktsar districts as P/PET > 0.5, farmers may start for land preparation and sowing.

• In Nawashahar, Sangrur, Patiala , Fatehgarh sahib and Ropar districts as P/PET < 0.5, Starts for land preparation. Wait for proper sowing period.

Hail storm likely in the districts of Himachal Pradesh (through Doppler Radar Observations)

Chamba

Kangra

Una

Solan

Sirmaur

Kinnaur

Shimla

Mandi

Kullu

Lahul and Spiti

Hamirpur

Bilaspur